The Fed Won’t Take Away the Punch Bowl, But the Party May Stop Anyway

We had a little bit of everything this week on Wall Street. An FOMC meeting and presser from newly appointed Kevin Warsh, earnings reports from some of the world’s largest companies transforming themselves from capital light free cash flow machines to capital intensive models with high customer concentration, and a flamboyant 25-year-old hedge fund manager with Silicon Valley ties gets margined out of his levered concentrated bets.

Kevin Warsh says the Fed is serious about its inflation mandate but is now engaging in can-kicking much like his predecessors. He wants to buy more time before deciding the direction of policy at the Fed to avoid a potential misstep. Warsh at one point said “Monetary policy matters, not just by what we say or even what we do. Monetary policy matters by how it affects the real economy”. Despite leaving rates unchanged, the market reacted by selling treasuries, and to Warsh’s point the real economy was impacted with mortgage rates reaching 6.66% later in the week. But consumers have grown insensitive to policy and longer-term rates; nobody cancelled their weekend plans because the 30yr treasury yield settled at 5.28%. Not even prospective first-time home buyers would be reconsidering, to the extent that they can still get a mortgage. The consensus among the public is that home prices only move higher and rates are what they are (people need to live) so the rise in yields at the long end won’t do much.

The Fed’s chosen utilization of its toolkit has grown ineffective. Unless Warsh pivots to emulate something closer to Volcker rather than this ill-fated attempt at can-kicking resembling the Greenspan era, it is the wealth effect that has the reins of the American economy. Put simply: Nothing the Fed can do outside of going Volcker can stop inflation or the wealth-effect. You can’t have one and not the other; they’re inseparable at this point. But the central bank isn’t the only market force, and of course Warsh and FOMC members are wary of the wealth effect. Surely, they are observing the current situation on Wall Street as closely as all of us (which would help to explain, in part, their wait and see approach). A handful of the world’s largest companies, representing a material percentage of the index centered at the core of the equity portfolio in millions of 401(k) and retirement savings accounts, are now pricing in extraordinary returns on a new and highly speculative industry in a material way, and have in effect ushered in a brand-new era of big-tech counterparty risk.

The index prescribed for the defensive investor by Ben Graham and John Bogle many decades ago is now highly concentrated in a handful of unfathomably large technology companies that are amidst a capital-intensive transformation in the magnitude of trillions by 2030. Most of these companies also have an unusually large share of contracted future revenue sitting with two counterparties that have no public financials. OpenAI and Anthropic are burning cash despite peak excitement and dominant market share and are facing new competition from a stream of cheaper and open-source alternatives. A stream that is just starting and will perhaps turn into an ocean of alternatives. Their contracts with big tech (and subsequent impact to valuations), via the wealth effect, essentially have more influence over the American economy than the central bank due to the latter’s unwillingness to intervene appropriately. Even more concerning to me is that much like the promises of the metaverse from only a few years ago, consumers don’t even have great use cases, and the enterprise business is extremely new. It still rests on some frankly bold assumptions about the willingness by businesses to adopt the technology into corporate operations: the reality is that these companies have their own IT solutions, their own security protocols, and are not about to hand the enterprise keys to the companies that are self-reporting the cybersecurity messes that they’re casually creating. It feels like the elephant in the room…what would happen if OpenAI failed? The intelligent investor must ask this question now.

I think that Sam Altman and the entire industry (intentionally or not) orchestrated a “too big to fail” technology boom that’s roped in the same companies that have driven the bull market post GFC, who had perhaps the greatest business models in human history, converting them into capital intensive business models that are financing their largest customers. If a report should surface that investors interpret as even an inkling of an admission that OpenAI’s business model is getting to be unsustainable, then the corresponding revenue growth, earnings, and free cash flow being priced in for the world’s largest companies would come under pressure in a material way. A report like that, I believe, would not come as a shock to read (I’m old enough to remember FTX! How about that?). It would introduce extreme market volatility and begin to materially diminish the wealth effect’s impact on consumer spending and the real economy. The financing is not a characterization. Amazon has invested $50.0 billion into OpenAI, fully funded, against OpenAI’s $138.0 billion purchase commitment to AWS; it has prepaid more than a third of its own contract back to its customer as equity. Amazon has also made available to Anthropic a facility “not to exceed $20.0 billion” where, in its own words, “as we reach certain delivery milestones of compute capacity … amounts under this facility are made available for Anthropic to draw upon.” The seller’s credit to the buyer unlocks as the seller ships. Across the whole complex, $61.9 billion of confirmed cash has gone into OpenAI against $706.4 billion of published commitments coming back the other way.

I wrote in Weeds and Flowers of the importance for hyper-scaler capex to be term-outs of strength. Oracle spent $55.66 billion on capital expenditures this fiscal year, up 162%, funded with $42.7 billion of notes and $5.0 billion of preferred stock against $32.0 billion of operating cash flow. That is not a term-out of strength. Meta is the other half of this problem, and in one respect the worse half: it holds no lab contract at all. Microsoft and Oracle at least carry a signed obligation from a counterparty; Meta is spending against demand it has only forecast for itself. Despite META’s wonderful advertising business, it now faces a widening funding gap, and the numbers are its own. Capital expenditures of $31.08 billion in the quarter against $31.86 billion of operating cash flow left free cash flow of $784 million, where a year earlier it was $8.5 billion. Buybacks went to zero, from $10.2 billion in the same quarter last year. Long-term debt rose to $83.7 billion from $58.7 billion at year end, with $24.9 billion issued in the quarter alone. The 10-Q from this week failed to offer sufficient insight into their AI business’s prospects. Investors are willing to wait for what they’re being promised, but who can’t help but get flashbacks to only a few years ago when the same company made promises about a “metaverse” that never came to fruition. Perhaps a bit of Monday morning quarterbacking here, but I think the median consumer with median intellect could have warned Mark Zuckerberg that he was misallocating capital with that one, but many CEOs, to me, are seemingly out of touch. The same thing is true here. Some could just be poor stewards of capital and are not fit for capital allocation, which is half the job especially after creating a FCF printer like Zuckerberg has built. But I don’t want to pick on META which everyone is doing because it’s not just one company that may be misallocating capital in a historically bad way. Investors point to AWS, Azure and Google Cloud’s early returns as validation. What they cannot see is how much of that revenue is circular. Only Microsoft has had to tell us, and only because the accounting rules forced it: $24.1 billion of fiscal 2026 revenue from commercial arrangements with OpenAI, a company in which Microsoft holds roughly a 25% stake and to which it has funded $11.9 billion of a $13.0 billion commitment. Oracle, whose remaining performance obligations went from $138 billion to $638 billion in a single year, does not name a customer anywhere in its 10-K.

The degree of certainty at which executive management and the technology industry broadly are operating under should not provide comfort to the intelligent enterprising investor, and, if anything, should raise more questions. Not only are the expected revenues attributed to planned capex ipso facto speculative, but the productivity gains promised by artificial intelligence are being oversold: the gains large enough to meet expectations come from massive reductions to corporate headcount across all companies in all industries, to improve operating leverage. But at what cost to the overall economy? Reduced employment and consumer spending would offset the economic benefits of increased operating leverage. It’s under-appreciated that these capex plans are speculative bets on the overall demand for this new technology. Nvidia’s own 10-K says its customers “can generally cancel, change, or delay” orders “with little notice to us and without penalty,” while Nvidia itself carries $119 billion of supply and capacity commitments, $95 billion of it payable within the coming fiscal year. The company with the largest revenue in this build-out holds the weakest contracts in it. That’s just a fact. An investment promises safety of principal and adequate return, both of which remain at large here because the industry is not yet established. These companies were capital-light compounders, compounding at rates never seen before at their size. Free cash flow printers. Alphabet reported diluted earnings of $9.11 per share for the June quarter, up 294%. Its own release attributes $6.26 of that to a non-cash gain on equity securities. In the same quarter it produced negative $5.9 billion of free cash flow, repurchased no stock at all against $13.2 billion a year earlier, and raised $49.6 billion of equity and $20.3 billion of notes. This seems to have been taken for granted by their own management as they rushed into something so new in such a material and aggressive way that they’re now capital intensive.

These managers could learn a great deal from the greatest steward and allocator of capital that’s ever lived. Warren Buffett’s Berkshire Hathaway has a record cash pile waiting to be allocated and has now for several years leading up to his retirement. But even as he faced retirement and is now approaching the end of his life, this cash has sat idly. Does he think all businesses today just suck? Of course not, it’s that his limited universe of opportunities has yet to produce the promise of safe principal and adequate return at the scale that is required at Berkshire. Nothing will infiltrate his discipline as an allocator & steward of capital. Nobody more than Warren wants to make one last giant whale of an acquisition, but he would rather retire than invest unintelligently. It’s an under-discussed, under-appreciated, and extremely powerful message that will last for generations of investors to learn from. Some of the world’s largest companies have demonstrated ignorance of these principles.